Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Difference between two age-specific incidence rates and IRR following Poisson regression

From   Tim Evans <>
To   "" <>
Subject   st: Difference between two age-specific incidence rates and IRR following Poisson regression
Date   Mon, 1 Oct 2012 11:39:39 +0100

Hi all,

I hope you may be able to help.

I am looking at the age specific incidence rates of a type of surgery in cancer patients in a given age group over set periods in time. I have then run a simple Poisson regression model to assess the incidence rate ratio between a base period and following periods in time using the same age group.

So for instance.

One set of data I find the following:

major surgery rate per 100,000 people in 1988-89 is 73.5 (in the 50-59 age group)
major surgery rate per 100,000 people in 2008-10 is 96.1 (in the 50-59 age group)

If I calculate the rate ratio of 2008-10 to 1988-89 from the data above I have a ratio of 1.31

If I calculate the IRR using Poisson regression for the same data, I return a value of 0.86 (CIs 0.79, 0.93) - which I am clearly finding hard to explain.

The age-specific incidence rates are calculated by dividing the count of surgeries in the given population group and multiplied by 100,000. The poisson regression code in Stata looks like this:

poisson maj_surg flag2 if age=="50-59", irr

maj_surg is binary 0 = no major surgery, 1 = major surgery
flag5 is binary 0 = 1988-89, 1 = 2008-10

The major surgery rates calculated above are age specific calculations, whereas the IRR calculated (although based upon the same age groups) are not necessarily, but as I'm presenting age specific results, this is surely not an area of difference.

Are the difference related to the fact that my age-specific rates do not take account of any underlying distribution of the data whereas I am implicitly saying in the Poisson regression calculation that the data have a Poisson distribution, and as such the underlying calculations are going to be different?

Any pointers really appreciated.

Best wishes


This email and any attachments hereto contains proprietary information, some or all of which may be confidential or legally privileged. It is for the exclusive use of the intended recipient(s) only. If an addressing or transmission error has misdirected this e-mail and you are not the intended recipient(s), please notify the author by replying to this e-mail. If you are not the intended recipient you must not use, disclose, distribute, copy, print, or rely on this e-mail or any attachments, as this may be unlawful.

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index